dataset_list.py

from __future__ import annotations

import numpy as np
import pandas as pd
import requests
from huggingface_hub.hf_api import SpaceInfo

url = 'https://docs.google.com/spreadsheets/d/1RoM2DgzaYJg6Ias1YNC2kQN01xSWJb1KEER9efb0X7A/edit#gid=0'
csv_url = url.replace('/edit#gid=', '/export?format=csv&gid=')

class DatasetList:
    def __init__(self):
        self.table = pd.read_csv(csv_url)
        self._preprocess_table()

        self.table_header = '''
            <tr>
                <td width="15%">Dataset Name</td>
                <td width="10%">Question Type</td>
                <td width="10%">Applied In Paper</td>
                <td width="10%">Reference Paper</td>
                <td width="20%">Brief Description</td>
                <td width="5%">Count</td>
                <td width="10%">Original Access Link</td>
                <td width="10%">Publicly Available?</td>
                <td width="10%">Access link on 🤗</td>   
            </tr>'''

    def _preprocess_table(self) -> None:
        self.table['dataset_name_lowercase'] = self.table.dataset_name.str.lower()
        self.table['count'] = self.table['count'].apply(str)

        rows = []
        for row in self.table.itertuples():
            dataset_name = f'{row.dataset_name}' if isinstance(row.dataset_name, str) else ''
            question_type = f'{row.question_type}' if isinstance(row.question_type, str) else ''
            used_in_paper = f'{row.used_in_paper}' if isinstance(row.used_in_paper, str) else ''
            reference_paper = f'<a href="{row.reference_paper}" target="_blank">Paper</a>' if isinstance(row.reference_paper, str) else ''
            brief_description = f'{row.brief_description}' if isinstance(row.brief_description, str) else ''
            count = f'{row.count}' if isinstance(row.count, str) else ''
            original_link = f'<a href="{row.original_link}" target="_blank">Access Link</a>' if isinstance(row.original_link, str) else ''
            publicly_available = f'<a href="{row.publicly_available}" target="_blank">License</a>' if isinstance(row.publicly_available, str) else ''
            huggingface_link = f'<a href="{row.huggingface_link}" target="_blank">HF Link</a>' if isinstance(row.huggingface_link, str) else ''
            row = f'''
                <tr>
                    <td>{dataset_name}</td>
                    <td>{question_type}</td>
                    <td>{used_in_paper}</td>
                    <td>{reference_paper}</td>
                    <td>{brief_description}</td>
                    <td>{count}</td>
                    <td>{original_link}</td>
                    <td>{publicly_available}</td>
                    <td>{huggingface_link}</td>
                </tr>'''
            rows.append(row)
        self.table['html_table_content'] = rows

    def render(self, search_query: str, 
            case_sensitive: bool,
            filter_names: list[str]
            ) -> tuple[int, str]:
        df = self.table
        if search_query:
            if case_sensitive:
                df = df[df.dataset_name.str.contains(search_query)]
            else:
                df = df[df.dataset_name_lowercase.str.contains(search_query.lower())]
        has_dataset = 'Dataset' in filter_names
        has_datalink = 'Data Link' in filter_names
        has_paper = 'Paper' in filter_names
        df = self.filter_table(df, has_dataset, has_datalink, has_paper)
        #df = self.filter_table(df, has_paper, has_github, has_model, data_types, model_types)
        return len(df), self.to_html(df, self.table_header)

    @staticmethod
    def filter_table(df: pd.DataFrame, 
                     has_dataset: bool, 
                     has_datalink: bool,
                     has_paper: bool
                    ) -> pd.DataFrame:
        if has_dataset:
            df = df[~df.dataset_name.isna()]
        if has_datalink:
            df = df[~df.huggingface_link.isna() | ~df.original_link.isna()]
        if has_paper:
            df = df[~df.reference_paper.isna()]
        # df = df[df.data_type.isin(set(data_types))]
        #df = df[df.base_model.isin(set(model_types))]
        # df = df[df.year.isin(set(years))]
        return df

    @staticmethod
    def to_html(df: pd.DataFrame, table_header: str) -> str:
        table_data = ''.join(df.html_table_content)
        html = f'''
        <table>
            {table_header}
            {table_data}
        </table>'''
        return html